1. Field of the Invention
This invention relates to a predictive coding system, and more particularly to a predictive coding system which handles data having a high transmission rate.
2. Description of the Prior Art
As means for coding when signals having temporal correlation such as video signals and audio signals are digitally transmitted, there is known predictive coding such as the well-known differential PCM coding (hereinafter termed DPCM). For example, when signals having a large amount of information such as television signals are transmitted, it is necessary to decrease bit rates down to transmission bit rates which conform to a transmission line. It can be considered to use predictive coding as means for that purpose.
Now, it is necessary to perform predictive coding processing with a high speed when the transmission rate is extremely high, but there exists a limitation in realizing high-speed processing.
There is generally known, as means for realizing a high-speed digital data processing, means wherein each of the data of a data sequence is sequentially and cyclically supplied to a plurality of processing systems for each sample, and the processing speed of each system is decreased down to a fraction of plural numbers to increase the processing speed as a whole.
Accordingly, when this means is applied to predictive coding processing, a high-speed processing can be realized.
However, when the above-described means is applied to the predictive coding processing of image data, for evaluating the predictive value of a picture element in each system, i.e., each coder, the data input immediately before, i.e., the data of an adjacent picture element, is not utilized. For example, when one-dimensional DPCM coding of video data transmitted in raster is performed, a predictive value is produced with a picture element separated by a plurality of picture elements in the horizontal direction on a picture, so that predictive error becomes large. This is because the correlation between a picture element used for producing the predictive value of the picture element in question, and the latter picture element, decreases.
In general, differential data are non-linearly quantized for the purpose of reducing the amount of data. When the value of predictive error is large, the difference between the typical value of non-linearly quantized data and the true value becomes large, resulting in deterioration of image data transmitted.
As for the predictive coding of an image, it is possible to use the so-called two-dimensional predictive coding which uses the correlation in the vertical direction of an image, as well, to further reduce the predictive error. However, when parallel processing of data is performed by using a plurality of predictive coders, an image used for prediction can not be freely selected in the case of performing the aforementioned processing, so that it is impossible to reduce the predictive error.
Further, in the case of DPCM coding of color video signals, the sampling frequency of color signals is generally set lower than the sampling frequency of luminance signals because details of the color attract less of the viewer's attention. In this case, two DPCM-coding circuits and clock circuits for driving are required, one of each for luminance signals and one of each for color signals.